5-Min Brief: The US and China Are About to Talk About AI at the Highest Level. Here's Why That Matters.

What you need to know — in 30 seconds

• Trump travels to Beijing on May 14–15 for a summit with Chinese President Xi Jinping
• The Wall Street Journal reports AI is being considered for the agenda — which would mark the first formal US-China AI dialogue under the Trump administration
• The proposed talks would focus on risks from unpredictable AI models, autonomous weapons, and the threat of non-state actors using advanced AI
• Expectations from analysts are low — but even opening a formal communication channel would be significant

The two countries that dominate AI development have barely talked to each other about it at the official level. That may be about to change.

President Trump heads to Beijing next week for his most significant summit with Chinese President Xi Jinping since taking office. Trade, Taiwan, and the ongoing Iran conflict are expected to dominate the agenda. But according to the Wall Street Journal, AI is being quietly added to the mix — a move that analysts are describing as potentially the most consequential item on the list, even if it generates the fewest headlines.

Why US and China need to talk about AI

Here's the core problem. The United States and China are both racing to build the most powerful AI systems in the world. Both are integrating AI into military systems, critical infrastructure, and strategic decision-making. And right now, they have almost no formal communication channels to manage what happens if something goes wrong.

Think about what that means in practice. If an AI system deployed by one country behaves in an unexpected way that the other country interprets as a threatening action — a misread sensor, an autonomous system acting outside expected parameters, a cyberattack that looks like it came from an AI but didn't — there's no hotline, no agreed protocol, no shared framework for what happens next.

That's the kind of gap that historically gets people killed. Nuclear weapons created the same problem in the 1950s and 60s, and the solution — arms control treaties, direct communication lines, shared safety norms — took decades of painful diplomacy to build. AI is moving faster than nuclear weapons did. The diplomatic infrastructure isn't keeping up.

What the proposed talks would actually cover

The proposed AI dialogue would focus on three specific areas, according to sources cited by the Wall Street Journal.

First: risks from unpredictable AI model behavior. As AI systems become more capable and are deployed in higher-stakes environments, the possibility of unexpected failures or outputs becomes a genuine concern. Both governments are aware that models can behave in ways their creators didn't anticipate.

Second: autonomous weapons systems. Both the US and China are developing AI-powered military systems — drones, targeting systems, logistics. The question of how much autonomous decision-making authority these systems should have, and what guardrails exist, is one of the most sensitive and consequential questions in modern warfare.

Third: non-state actors using advanced AI. As powerful AI models become more widely available — through open-source releases and leaked weights — the risk that terrorist groups, criminal organizations, or rogue states could use them for large-scale harm is a shared concern that cuts across the competitive dynamic between the two countries.

Treasury Secretary Scott Bessent is reportedly leading the American side of the proposed AI discussions. Beijing has not yet named its counterpart — a detail worth watching, since who China assigns to these talks will signal how seriously it's taking them.

Why the last attempt didn't work

The US and China actually did open formal AI talks under President Biden. They started at a California summit in 2023 and established an ongoing dialogue framework. The results were limited.

The reason, according to Georgetown scholar Rush Doshi — who led those talks on the US side — was that China assigned its foreign ministry to run the negotiations rather than technical experts. Diplomats and technical AI researchers are trying to solve fundamentally different problems. Diplomats manage relationships and avoid conflict. Technical experts need to understand the actual mechanics of how AI systems fail. Mixing them up produces talks that look productive but don't actually reduce risk.

Doshi told the Wall Street Journal the US should have pushed harder for deeper technical engagement. The question for next week's summit is whether either side has learned that lesson.

What this means for the bigger picture

We covered the Stanford AI Index last month, which noted that the US-China AI race is considerably closer than most headlines suggest. Chinese models from DeepSeek and Alibaba trail the best American models by only a few percentage points on key benchmarks. Both countries are spending heavily on AI infrastructure, though the US leads significantly on absolute spending.

That competitive dynamic makes formal dialogue both more important and more difficult. More important because the stakes of miscalculation are higher when both sides are genuinely capable. More difficult because neither side wants to share information that might help the other close the gap.

The optimistic scenario from next week's summit: both sides agree to establish a regular communication channel, share limited information about AI safety incidents, and begin developing non-binding norms around autonomous weapons. None of that solves the underlying competition. But it reduces the chance that competition tips into catastrophe.

Analysts from Brookings, Carnegie, and other institutions are urging low expectations — and that's probably the right posture. But even a phone call between technical experts would be more than exists today.

HumanReadable-AI covers AI news in plain English every weekday. Subscribe below — free, no jargon, always under five minutes.